Capacitor in parallel with motor - What happens if I remove it?

[paul8f]

Hi all,

see schematic diagram.

Just wondering what the function of C95 might be, and what happens if I leave it out of the circuit?

Paul.

[m3vuv]

i guess its there to cut emc/hash interfereance?

[bdunham7]

What is the signal that drives the FAN input (Q50 base)?  If it is PWM, then the capacitor integrates (smooths) that to a steady voltage. 

[paul8f]

i guess its there to cut emc/hash interfereance?

I've never heard of hash interference before. What kind of interference is that?

What is the signal that drives the FAN input (Q50 base)?

Hi. The signal comes from a PIC controller, and the fans can be heard changing RPM at times; so yeah I'm guessing it's PWM. (I will put the scope on it later today when I'm home from work to verify this...).

[mikerj]

If it's driven with PWM then that capacitor should not be there at all, it will be pulling large current transients from the supply rail and through the transistor.

[wraper]

What is the signal that drives the FAN input (Q50 base)?  If it is PWM, then the capacitor integrates (smooths) that to a steady voltage. 

If you are supplying brushless DC fan with PWMed power, you are an idiot. If you "smooth" PWM by capacitor without what basically is a buck converter, you are an idiot as well. Such fans contain electronics and often some capacitor as well. They consume power in pulses since windings are switched and capacitor helps keeping the power stable.

[wraper]

As of circuit itself, either it's simply used to switch the fans ON/OFF with no speed control or it should be broken otherwise. It's not an emitter follower, nor there is any feedback. Thus it's basically not possible to properly control voltage or current at it's output if used in linear mode. But if it used PWM, then someone who designed that would be a complete dumbass.

[bdunham7]

Wow!  So much angst over the simplest of circuits.   

Simple, low parts count, BOM less than $1.00 even in quantity 1 at retail, a fraction of that in bulk.  Probably works perfectly in the device within which it is installed.  At least as efficient as a linear driver would be.  2SD882 is plenty robust enough to handle any transients without additional components.  b/t/w, surge currents through the capacitor are not a problem, they are how these types of circuits actually work.  The only interesting component is the FR107.  I personally might add one resistor--can anyone guess where?--but it probably isn't needed in the application it is in.

OP:  Is there an issue with the circuit or a reason you want to leave the capacitor out?

[wraper]

Simple, low parts count, BOM less than $1.00 even in quantity 1 at retail, a fraction of that in bulk.  Probably works perfectly in the device within which it is installed.  At least as efficient as a linear driver would be.  2SD882 is plenty robust enough to handle any transients without additional components.  b/t/w, surge currents through the capacitor are not a problem, they are how these types of circuits actually work.  The only interesting component is the FR107.  I personally might add one resistor--can anyone guess where?--but it probably isn't needed in the application it is in.

If you PWM it, quite soon that 10uF capacitor will fail and probably even blow up including capacitor in the fan if there is such, fan will start making strange noises and could fail as well.

[bdunham7]

If you PWM it, quite soon that 10uF capacitor will fail and probably even blow up including capacitor in the fan if there is such, fan will start making strange noises and could fail as well.

And you know all that despite not knowing the characteristics of the fans, the voltage and ripple rating of the capacitor nor the ambient temp this runs at--and despite the fact that the OP has not yet stated any problem with the device? 

[schmitt trigger]

My personal experience on PWM'ing brushless fans. Which may or may not apply to your question.

Many years ago, I worked in a large server building. With so much processing power packed tightly, one of the major problems is removing the heat.
All of the cooling equipment, which includes the chillers, cold water pumps, fans and ancillary equipment consumes several hundred kilowatts, a fantastic amount.
Thus controlling and optimizing the cooling requirements is of paramount importance.

For the purpose of this discussion, I am focusing on the fans that exhausted the heat from a server rack. Each rack had two large 60 VDC fans, which were of course PWM'ed based on exhaust air temperature.
There were close to a thousand such fans, operating 24/7. Because of the reliability requirements they were premium grade devices with ball bearings. No cheap bushings.

We would get 2 or 3 temperature alarms per week. It was always the same fault, the fan was turning slowly. And the reason was always the same, the ball bearings had degraded causing the fan to run slow even at full voltage.

Long story short: Soon replacing the fans became very expensive and therefore we contacted the manufacturer.

The very first question from them: Are you PWM'ing the fans?  If so, this PWM action causes internal forces which wear out the bearings prematurely.

Since we had to control the fan speed, our solution was to put an electrolytic cap in parallel with each fan. That solved the problem.

I don't recall the exact value, it could have been 47uF. What I do recall is that we measured the ripple current through the cap, and we selected a capacitor from a reputable manufacturer, and we selected a series which would withstand high ripple currents. Not all electro caps are created equal.

[wraper]

If you PWM it, quite soon that 10uF capacitor will fail and probably even blow up including capacitor in the fan if there is such, fan will start making strange noises and could fail as well.

And you know all that despite not knowing the characteristics of the fans, the voltage and ripple rating of the capacitor nor the ambient temp this runs at--and despite the fact that the OP has not yet stated any problem with the device? 

I know that ripple current rating of such capacitor is a few hundreds of mA at best. Given it's an ultra super duper LOW ESR type.

[bdunham7]

I know that ripple current rating of such capacitor is a few hundreds of mA at best. Given it's an ultra super duper LOW ESR type.

LOW ESR?  Do you mean the LOWESR brand?



Yes, if it is a smallish, cheap el-cap then that's right.  But how big are the fans?  Many of these small box fans are only 50-200mA @12V and that goes down as the duty cycle and voltage drop.

I've no idea why the OP is asking.  Perhaps it is because the 10uF cap keeps exploding and he wants to leave it out.  Improper implementation doesn't mean the basic principle is wrong.  If the fans are bigger, than a more robust cap could be needed.

https://www.grainger.com/product/DAYTON-Oval-Motor-Run-Capacitor-2MDV7?searchBar=true&searchQuery=2MDV7

[wraper]

LOW ESR?  Do you mean the LOWESR brand?

I'm not sure if you are serious or joking. BTW cap on the photo does not have LOWESR brand.

[paul8f]

Yup, here's that signal at the base of Q50. Difficult enough to lock onto it using an old analogue scope.

The reason I'm asking about C95, is because this cap cracked and shorted out, blowing Q51 in the process. Fans worked again after C95 removed, and Q51 replaced.

But when I went looking for a replacement 10uF cap, I couldn't find an SMD electrolytic in a 1206 size. I ended up getting this multi-layer ceramic cap from Farnell instead:   https://ie.farnell.com/tdk/c3216x7r1e106k160ab/cap-10-f-25v-10-x7r-1206/dp/2525177?ost=2525177.

It's not polarized but no issues since I soldered it in.

Anyway, this is a cheap Chinese inverter, and has worked well for many years. Now I'm trying to get a few more out of it!

[bdunham7]

LOW ESR?  Do you mean the LOWESR brand?

I'm not sure if you are serious or joking. BTW cap on the photo does not have LOWESR brand.

Well, the capacitor in the photo is neither Low ESR nor Vented, since the bottom popped out instead.  So I figured that must be the brand. 

[Gyro]

By coincidence, I found exactly the same design issue on a GW Instek GPS2303 bench PSU - also a switching transistor and 10uF capacitor driven by a basic PWM signal.

I fixed it sufficiently for my needs by converting it into a fairly crude RC filter - not enough to produce a smooth DC supply, but enough to keep the fan happy and quiet. It's still working smoothly 4 years later.

https://www.eevblog.com/forum/repair/cured-my-gps2032-psu-rattle/

[bdunham7]

Yup, here's that signal at the base of Q50. Difficult enough to lock onto it using an old analogue scope.

The reason I'm asking about C95, is because this cap cracked and shorted out, blowing Q51 in the process. Fans worked again after C95 removed, and Q51 replaced.

But when I went looking for a replacement 10uF cap, I couldn't find an SMD electrolytic in a 1206 size. I ended up getting this multi-layer ceramic cap from Farnell instead:   https://ie.farnell.com/tdk/c3216x7r1e106k160ab/cap-10-f-25v-10-x7r-1206/dp/2525177?ost=2525177.

It's not polarized but no issues since I soldered it in.

Anyway, this is a cheap Chinese inverter, and has worked well for many years. Now I'm trying to get a few more out of it!

Are you saying the original cap was 1206 and electrolytic?  That would, in fact, seem inadequate.  Your replacement may be OK, you can just check and make sure it doesn't get too hot in operation, specifically when the fans slow down from full speed.  The lack of polarization is not an issue.  What frequency is that signal?

Edit: The Kemet T527S106M025ATE150 is 1206, 10uF @25VDC, ESR=150mR, rated ripple current is 632mA @ +45C.  Presumably a lot less as it warms up.  So a 1206 package is feasible.  I guess you reached the end-of-life for whatever was in there.

[Gyro]

Without some resistance in circuit, the transistor is in imminent danger of failure again, particularly if the cap is low ESR or ceramic, due to the current peaks.

[paul8f]

Are you saying the original cap was 1206 and electrolytic?  That would, in fact, seem inadequate.  Your replacement may be OK, you can just check and make sure it doesn't get too hot in operation, specifically when the fans slow down from full speed.  The lack of polarization is not an issue.  What frequency is that signal?

Cap is not getting hot, and the PWM photo was taken at medium fan RPM. The frequency was 27.8kHz.

I'm thinking that if Q51 blows again, I'd never know, as the inverter only brings on a buzzer and warning LED for 2 conditions:  Current Over-load, and Over-Temperature safety shutdown. I could possibly add my own FAN TEST pushbutton switch to periodically test the fan circuit. (Connect Vcc through a Push-to-make switch, and a 1K resistor to the base of Q50). Push and listen for fan-spin

[bdunham7]

Without some resistance in circuit, the transistor is in imminent danger of failure again, particularly if the cap is low ESR or ceramic, due to the current peaks.

I'm not sure that's true.  The 1K resistor limits the base current of Q51 to 12mA so even with a gain of 100 it would still limit the current to 1.2A.  No?

[bdunham7]

Cap is not getting hot, and the PWM photo was taken at medium fan RPM. The frequency was 27.8kHz.

If it isn't too much trouble, can you scope the base of Q51 as well?  I'm not seeing a signal that would account for 'medium fan speed'.

[Gyro]

Without some resistance in circuit, the transistor is in imminent danger of failure again, particularly if the cap is low ESR or ceramic, due to the current peaks.

I'm not sure that's true.  The 1K resistor limits the base current of Q51 to 12mA so even with a gain of 100 it would still limit the current to 1.2A.  No?

Hard to know to be honest - the hfe of the 2SD882 is 100 min and 300 max. I usually find that transistors tend to be towards the upper end of their hfe range. If the transistor goes into current limit then it's Vce will rise (it would be able to go quite a long way with a 24V supply). Without careful analysis, I wouldn't be willing to guarantee the survival of the transistor, either through Ic or overall dissipation.

It seems an awful lot to be current spiking in that way, it's going to be electrically (and mechanically? noisy). You can see what I did on the Instek PSU anyway.

[bdunham7]

Without some resistance in circuit, the transistor is in imminent danger of failure again, particularly if the cap is low ESR or ceramic, due to the current peaks.

I'm not sure that's true.  The 1K resistor limits the base current of Q51 to 12mA so even with a gain of 100 it would still limit the current to 1.2A.  No?

Hard to know to be honest - the hfe of the 2SD882 is 100 min and 300 max. I usually find that transistors tend to be towards the upper end of their hfe range. If the transistor goes into current limit then it's Vce will rise (it would be able to go quite a long way with a 24V supply). Without careful analysis, I wouldn't be willing to guarantee the survival of the transistor, either through Ic or overall dissipation.

It seems an awful lot to be current spiking in that way, it's going to be electrically (and mechanically? noisy). You can see what I did on the Instek PSU anyway.

Well, then further consider that other than the few hundreds of mA in fan current, the 'spike' is through a 10uF cap--and that cap is only discharged fully at initial turn-on.  During normal operation I would expect the peak current to be about twice the average and no more due to the fact that the cap will still be almost at full voltage.  As for the turn-on spike, hFe will go down sharply at higher currents, but the transistor is rated for 6A peaks up to 5mS and would seem ti be limited to 3.6A even with a gain of 300.  It's worth analyzing, perhaps, if you wanted a reliable, quality product--but I don't think this is one.

Your Instek fan had a drive frequency of 65 Hz?   

[paul8f]

If it isn't too much trouble, can you scope the base of Q51 as well?  I'm not seeing a signal that would account for 'medium fan speed'.

No worries, see photos attached. I was thinking the same thing.... not much definition between the mark and the space.

In the third waveform, I'm still probing the base of Q51, but I've replaced the fans with a small 5w filament test-lamp. Then, when I put the inverter into an over-temp (safety shutdown) condition, the test-lamp stays illuminated as expected, but signal at the base at Q51 changes to a flat DC line on the scope (even though this is not a max RPM fan setting...). Still in safety shutdown mode, when I change from the test-lamp back to the fans, I see a pulse reappear at the base of Q51. So maybe my scope is picking up some interference from the fans, (or do these basic Jin Li fans have some internal components??  )

When I take the unit from safety shutdown mode back to normal run mode, the signal gets more complex again. So it seems I'm also picking up something from the inverter switching circuit too. My probe is in good condition, and I'm hooking the reference lead (gnd clip) up to the correct ground line very close to the area I'm probing.

It doesn't help that I'm so accustomed to hitting the 'Autoset' button on my scope at work.